Low reliability of the study

We can provide an independent, transparent data analysis process.

We can audit the study - procedure and statistical analysis.

We will provide a blinded statistician to generate a script for data analysis without access to information about which drug or therapy is assigned to which patient. 

due to lack of independent analysts, and blinded randomization. Data is collected by the same group that conducts the analysis, and full access to information on what drug or treatment is given. 

Incorrectly counted sample size

Calculating an appropriate sample size is an essential part of any clinical trial.

We provide a professional statistician to determine the sample size according to the study design, study purpose, and assumptions like

- Distribution of data
- Correlation of observations
- Correlation of endpoints
- Subgroup analysis including propensity score.  

due to failure to consider all survey assumptions or the use of overly strong assumptions (e.g., normal distribution of studied characteristics).  It is not uncommon for studies to be underpowered and thereby fail to detect the existing treatment effects due to inadequate sample size 

Improperly designed CRF

Before starting the study, we provide a review of the CRF template in terms of statistics, endpoints, and whether they can be counted from the data collected from the CRF 

resulting in incorrect or incomplete data acquisition.  For example, lack of data for endpoint statistical analysis.

Problems with data quality during the study

We provide ongoing statistical monitoring.

Two types of rules are implemented:

Hard rules - if the Protocol Deviation increases then corrective procedures are put in place - sending a monitor to the site to find the cause.

Soft rules - which fall within the standard deviation and allow further investigation. 

especially in multicenter trials.  Noting more protocol deviations at some centers 

Data mess

We ensure data cleaning.

We provide a three-stage process, involving repeated cycles of screening, diagnosing, and editing of suspected data abnormalities.

Screening:
Lack/Excess of data
Outliers / Inconsistencies
Strange patterns
Suspect analysis result

Diagnosis:
Errors, missing data
True extreme
True normal

Editing:
Correction
Deletion
Leave unchanged 

Missing data, jumbled data types, type errors, no conditional rules, incorrect naming 

Incorrect presentation of results

We can help you understand publications using advanced statistics, as well as perform complex data analysis.

We can quickly and reliably perform the required calculations, as well as provide a professional presentation of data and information in tables and figures.

We assist in formatting and submitting manuscripts to scientific journals, paying special attention to table and figure formats. 

including poorly designed tables, charts
Incorrect interpretation of results, drawing incorrect conclusions 

Problem with reproducing the results

We develop scripts in terms of individual protocol studies. The scripts provide automatic repeatable analysis of collected data. 

is especially common when the analysis is performed one at a time, without automated scripts, with descriptions of the various steps of data analysis missing 

Lack of adequate data quality 

We provide assistance in writing data acquisition protocols. 

often in studies, data are excluded due to poor quality, which may be due to the lack of clear regulations for data availabilities 

Lack of reproducibility of object segmentation 

We develop automatic or semi-automatic segmentation instead of manual segmentation, considering:

Repeatability. This means that when it is used on the same data, the result will not change.  

Consistency. The algorithm actually solves the given problem and performs the task, rather than doing something that is not important. For example, the algorithm can detect the sick part in all the different scans.

Accuracy, it is very important that the algorithm detects the diseased part as precisely as possible. Only with accurate data can accurate results be obtained.

Time efficiency. Results should be generated as quickly as possible so that the entire diagnostic process can also be accelerated. 

in medical images due to differences in methodological approaches, as well as the experience of medical experts     Time-consuming manual analysis 

Time-consuming manual analysis 

We estimate the correct size of the data set, along with writing a procedure for dividing it into training, and validation sets, taking into account the characteristics of the images and the changes in the objects in them.

So that the individual changes have a similar distribution in each set. 

Class disparities and over-fitting of predictive models under development